Learning Objectives:

Manipulating and Visualizing Data Frames

Last week you started to manipulate data tables (under the class of "data.frame" objects) using bracket notation, dat[ , ], and the dollar operator, dat$name, in order to select specific rows, columns, or cells. In addition, you have been creating charts with functions like plot(), boxplot(), and barplot(), which are part of the "graphics" package.

In this lab, you will start learning about other approaches to manipulate tables and create statistical charts. We are going to use the functionality of the package "dplyr" to work with tabular data in a more consistent way. This is a fairly recent package introduced a couple of years ago, but it is based on more than a decade of research and work lead by Hadley Wickham.

Likewise, to create graphics in a more consistent and visually pleasing way, we are going to use the package "ggplot2", also originally authored by Hadley Wickham, and developed as part of his PhD more than a decade ago.

Use the first hour of the lab to get as far as possible with the material associated to "dplyr". Then use the second hour of the lab to work on graphics with "ggplot2" (all the material is in this file).

While you follow this lab, you may want to open these cheat sheets:

Installing packages

I’m assuming that you already installed the packages "dplyr" and "ggplot2". If that’s not the case then run on the console the command below (do NOT include this in your Rmd):

# don't worry too much if you get a warning message
install.packages(c("dplyr", "ggplot2"))

Remember that you only need to install a package once! After a package has been installed in your machine, there is no need to call install.packages() again on the same package. What you should always invoke in order to use the functions in a package is the library() function:

# don't forget to load the packages
library(dplyr)
library(ggplot2)

About loading packages: Another rule to keep in mind is to always load any required packages at the very top of your script files (.R or .Rmd or .Rnw files). Avoid calling the library() function in the middle of a script. Instead, load all the packages before anything else.

NBA Players Data

The data file for this lab is the same you used last week: nba2017-players.csv, which is located in the data/ folder of the course github repository. I assume that you already downloaded a copy of the csv file to your computer. If that is not the case, here’s one option to get your own copy:

# download RData file into your working directory
github <- "https://github.com/ucb-stat133/stat133-fall-2017/raw/master/"
csv <- "data/nba2017-players.csv"
download.file(url = paste0(github, csv), destfile = 'nba2017-players.csv')

To import the data in R you can use the base function read.csv(), or you can also use read_csv() from the package "readr":

# with "base" read.csv()
dat <- read.csv('nba2017-players.csv', stringsAsFactors = FALSE)

# with "readr" read_csv()
dat <- read_csv('nba2017-players.csv')

Basic "dplyr" verbs

To make the learning process of "dplyr" gentler, Hadley Wickham proposes beginning with a set of five basic verbs or operations for data frames (each verb corresponds to a function in "dplyr"):

I’ve modified Hadley’s list of verbs a little bit:

Filtering, slicing, and selecting

slice() allows you to select rows by position:

# first three rows
three_rows <- slice(dat, 1:3)
three_rows
## # A tibble: 3 x 15
##          player  team position height weight   age experience
##           <chr> <chr>    <chr>  <int>  <int> <int>      <int>
## 1    Al Horford   BOS        C     82    245    30          9
## 2  Amir Johnson   BOS       PF     81    240    29         11
## 3 Avery Bradley   BOS       SG     74    180    26          6
## # ... with 8 more variables: college <chr>, salary <dbl>, games <int>,
## #   minutes <int>, points <int>, points3 <int>, points2 <int>,
## #   points1 <int>

filter() allows you to select rows by condition:

# subset rows given a condition
# (height greater than 85 inches)
gt_85 <- filter(dat, height > 85)
gt_85
##               player team position height weight age experience
## 1        Edy Tavares  CLE        C     87    260  24          1
## 2   Boban Marjanovic  DET        C     87    290  28          1
## 3 Kristaps Porzingis  NYK       PF     87    240  21          1
## 4        Roy Hibbert  DEN        C     86    270  30          8
## 5      Alexis Ajinca  NOP        C     86    248  28          6
##                 college  salary games minutes points points3 points2
## 1                          5145     1      24      6       0       3
## 2                       7000000    35     293    191       0      72
## 3                       4317720    66    2164   1196     112     331
## 4 Georgetown University 5000000     6      11      4       0       2
## 5                       4600000    39     584    207       0      89
##   points1
## 1       0
## 2      47
## 3     198
## 4       0
## 5      29

select() allows you to select columns by name:

# columns by name
player_height <- select(dat, player, height)

Your turn:

  • use slice() to subset the data by selecting the first 5 rows.
first_five <- slice(dat, 1:5)
first_five
## # A tibble: 5 x 15
##              player  team position height weight   age experience
##               <chr> <chr>    <chr>  <int>  <int> <int>      <int>
## 1        Al Horford   BOS        C     82    245    30          9
## 2      Amir Johnson   BOS       PF     81    240    29         11
## 3     Avery Bradley   BOS       SG     74    180    26          6
## 4 Demetrius Jackson   BOS       PG     73    201    22          0
## 5      Gerald Green   BOS       SF     79    205    31          9
## # ... with 8 more variables: college <chr>, salary <dbl>, games <int>,
## #   minutes <int>, points <int>, points3 <int>, points2 <int>,
## #   points1 <int>
  • use slice() to subset the data by selecting rows 10, 15, 20, …, 50.
fives <- slice(dat, seq(10, 50, 5))
fives
## # A tibble: 9 x 15
##             player  team position height weight   age experience
##              <chr> <chr>    <chr>  <int>  <int> <int>      <int>
## 1    Jonas Jerebko   BOS       PF     82    231    29          6
## 2     Tyler Zeller   BOS        C     84    253    27          4
## 3      Edy Tavares   CLE        C     87    260    24          1
## 4       Kevin Love   CLE       PF     82    251    28          8
## 5 Tristan Thompson   CLE        C     81    238    25          5
## 6  DeMarre Carroll   TOR       SF     80    215    30          7
## 7   Lucas Nogueira   TOR        C     84    241    24          2
## 8      Serge Ibaka   TOR       PF     82    235    27          7
## 9    Daniel Ochefu   WAS        C     83    245    23          0
## # ... with 8 more variables: college <chr>, salary <dbl>, games <int>,
## #   minutes <int>, points <int>, points3 <int>, points2 <int>,
## #   points1 <int>
  • use slice() to subset the data by selecting the last 5 rows.
last_five <- slice(dat, (length(dat$player)-4):length(dat$player))
last_five
## # A tibble: 5 x 15
##            player  team position height weight   age experience
##             <chr> <chr>    <chr>  <int>  <int> <int>      <int>
## 1 Marquese Chriss   PHO       PF     82    233    19          0
## 2    Ronnie Price   PHO       PG     74    190    33         11
## 3     T.J. Warren   PHO       SF     80    230    23          2
## 4      Tyler Ulis   PHO       PG     70    150    21          0
## 5  Tyson Chandler   PHO        C     85    240    34         15
## # ... with 8 more variables: college <chr>, salary <dbl>, games <int>,
## #   minutes <int>, points <int>, points3 <int>, points2 <int>,
## #   points1 <int>
  • use filter() to subset those players with height less than 70 inches tall.
lt70 <- filter(dat, height < 70)
lt70
##          player team position height weight age experience
## 1 Isaiah Thomas  BOS       PG     69    185  27          5
## 2    Kay Felder  CLE       PG     69    176  21          0
##                    college  salary games minutes points points3 points2
## 1 University of Washington 6587132    76    2569   2199     245     437
## 2       Oakland University  543471    42     386    166       7      55
##   points1
## 1     590
## 2      35
  • use filter() to subset rows of Golden State Warriors (‘GSW’).
gsw <- filter(dat, team=="GSW")
gsw
##                  player team position height weight age experience
## 1        Andre Iguodala  GSW       SF     78    215  33         12
## 2          Damian Jones  GSW        C     84    245  21          0
## 3            David West  GSW        C     81    250  36         13
## 4        Draymond Green  GSW       PF     79    230  26          4
## 5             Ian Clark  GSW       SG     75    175  25          3
## 6  James Michael McAdoo  GSW       PF     81    230  24          2
## 7          JaVale McGee  GSW        C     84    270  29          8
## 8          Kevin Durant  GSW       SF     81    240  28          9
## 9          Kevon Looney  GSW        C     81    220  20          1
## 10        Klay Thompson  GSW       SG     79    215  26          5
## 11          Matt Barnes  GSW       SF     79    226  36         13
## 12        Patrick McCaw  GSW       SG     79    185  21          0
## 13     Shaun Livingston  GSW       PG     79    192  31         11
## 14        Stephen Curry  GSW       PG     75    190  28          7
## 15        Zaza Pachulia  GSW        C     83    270  32         13
##                                  college   salary games minutes points
## 1                  University of Arizona 11131368    76    1998    574
## 2                  Vanderbilt University  1171560    10      85     19
## 3                      Xavier University  1551659    68     854    316
## 4              Michigan State University 15330435    76    2471    776
## 5                     Belmont University  1015696    77    1137    527
## 6           University of North Carolina   980431    52     457    147
## 7             University of Nevada, Reno  1403611    77     739    472
## 8          University of Texas at Austin 26540100    62    2070   1555
## 9  University of California, Los Angeles  1182840    53     447    135
## 10           Washington State University 16663575    78    2649   1742
## 11 University of California, Los Angeles   383351    20     410    114
## 12       University of Nevada, Las Vegas   543471    71    1074    282
## 13                                        5782450    76    1345    389
## 14                      Davidson College 12112359    79    2638   1999
## 15                                        2898000    70    1268    426
##    points3 points2 points1
## 1       64     155      72
## 2        0       8       3
## 3        3     132      43
## 4       81     191     151
## 5       61     150      44
## 6        2      60      21
## 7        0     208      56
## 8      117     434     336
## 9        2      54      21
## 10     268     376     186
## 11      18      20      20
## 12      41      65      29
## 13       1     172      42
## 14     324     351     325
## 15       0     164      98
  • use filter() to subset rows of GSW centers (‘C’).
gsw_centers <- filter(gsw, position=="C")
gsw_centers
##          player team position height weight age experience
## 1  Damian Jones  GSW        C     84    245  21          0
## 2    David West  GSW        C     81    250  36         13
## 3  JaVale McGee  GSW        C     84    270  29          8
## 4  Kevon Looney  GSW        C     81    220  20          1
## 5 Zaza Pachulia  GSW        C     83    270  32         13
##                                 college  salary games minutes points
## 1                 Vanderbilt University 1171560    10      85     19
## 2                     Xavier University 1551659    68     854    316
## 3            University of Nevada, Reno 1403611    77     739    472
## 4 University of California, Los Angeles 1182840    53     447    135
## 5                                       2898000    70    1268    426
##   points3 points2 points1
## 1       0       8       3
## 2       3     132      43
## 3       0     208      56
## 4       2      54      21
## 5       0     164      98
  • use filter() and then select(), to subset rows of lakers (‘LAL’), and then display their names.
lal <- filter(dat, team=="LAL")
lakers <- select(lal, player)
lakers
##               player
## 1     Brandon Ingram
## 2       Corey Brewer
## 3   D'Angelo Russell
## 4        David Nwaba
## 5        Ivica Zubac
## 6    Jordan Clarkson
## 7      Julius Randle
## 8    Larry Nance Jr.
## 9          Luol Deng
## 10 Metta World Peace
## 11        Nick Young
## 12       Tarik Black
## 13   Thomas Robinson
## 14    Timofey Mozgov
## 15       Tyler Ennis
  • use filter() and then select(), to display the name and salary, of GSW point guards
gsw_salary <- select(gsw, player, salary)
gsw_salary
##                  player   salary
## 1        Andre Iguodala 11131368
## 2          Damian Jones  1171560
## 3            David West  1551659
## 4        Draymond Green 15330435
## 5             Ian Clark  1015696
## 6  James Michael McAdoo   980431
## 7          JaVale McGee  1403611
## 8          Kevin Durant 26540100
## 9          Kevon Looney  1182840
## 10        Klay Thompson 16663575
## 11          Matt Barnes   383351
## 12        Patrick McCaw   543471
## 13     Shaun Livingston  5782450
## 14        Stephen Curry 12112359
## 15        Zaza Pachulia  2898000
  • find how to select the name, age, and team, of players with more than 10 years of experience, making 10 million dollars or less.
ten_years <- filter(dat, experience>=10)
ten_mil <- filter(ten_years, salary<=10000000)
ty_and_lttm <- select(ten_mil, player, age, team)
ty_and_lttm
##               player age team
## 1      Channing Frye  33  CLE
## 2      Dahntay Jones  36  CLE
## 3     Deron Williams  32  CLE
## 4        James Jones  36  CLE
## 5        Kyle Korver  35  CLE
## 6  Richard Jefferson  36  CLE
## 7      Jose Calderon  35  ATL
## 8     Kris Humphries  31  ATL
## 9      Mike Dunleavy  36  ATL
## 10   Thabo Sefolosha  32  ATL
## 11       Jason Terry  39  MIL
## 12        C.J. Miles  29  IND
## 13     Udonis Haslem  36  MIA
## 14        Beno Udrih  34  DET
## 15        Randy Foye  33  BRK
## 16        David West  36  GSW
## 17       Matt Barnes  36  GSW
## 18  Shaun Livingston  31  GSW
## 19     Zaza Pachulia  32  GSW
## 20         David Lee  33  SAS
## 21      Lou Williams  30  HOU
## 22      Trevor Ariza  31  HOU
## 23      Brandon Bass  31  LAC
## 24       J.J. Redick  32  LAC
## 25       Paul Pierce  39  LAC
## 26    Raymond Felton  32  LAC
## 27        Boris Diaw  34  UTA
## 28     Nick Collison  36  OKC
## 29        Tony Allen  35  MEM
## 30      Vince Carter  40  MEM
## 31     Jameer Nelson  34  DEN
## 32       Mike Miller  36  DEN
## 33      Devin Harris  33  DAL
## 34        J.J. Barea  32  DAL
## 35 Metta World Peace  37  LAL
## 36   Leandro Barbosa  34  PHO
## 37      Ronnie Price  33  PHO
  • find how to select the name, team, height, and weight, of rookie players, 20 years old, displaying only the first five occurrences (i.e. rows)
rookies <- filter(dat, experience==0)
twenty_yo <- filter(rookies, age==20)
young_guys <- select(twenty_yo, player, team, height, weight)
young_ffive <- slice(young_guys, 1:5)
young_ffive
## # A tibble: 5 x 4
##              player  team height weight
##               <chr> <chr>  <int>  <int>
## 1      Jaylen Brown   BOS     79    225
## 2    Henry Ellenson   DET     83    245
## 3 Stephen Zimmerman   ORL     84    240
## 4   Dejounte Murray   SAS     77    170
## 5    Chinanu Onuaku   HOU     82    245

Adding new variables: mutate()

Another basic verb is mutate() which allows you to add new variables. Let’s create a small data frame for the warriors with three columns: player, height, and weight:

# creating a small data frame step by step
gsw <- filter(dat, team == 'GSW')
gsw <- select(gsw, player, height, weight)
gsw <- slice(gsw, c(4, 8, 10, 14, 15))
gsw
## # A tibble: 5 x 3
##           player height weight
##            <chr>  <int>  <int>
## 1 Draymond Green     79    230
## 2   Kevin Durant     81    240
## 3  Klay Thompson     79    215
## 4  Stephen Curry     75    190
## 5  Zaza Pachulia     83    270

Now, let’s use mutate() to (temporarily) add a column with the ratio height / weight:

mutate(gsw, height / weight)
## # A tibble: 5 x 4
##           player height weight `height/weight`
##            <chr>  <int>  <int>           <dbl>
## 1 Draymond Green     79    230       0.3434783
## 2   Kevin Durant     81    240       0.3375000
## 3  Klay Thompson     79    215       0.3674419
## 4  Stephen Curry     75    190       0.3947368
## 5  Zaza Pachulia     83    270       0.3074074

You can also give a new name, like: ht_wt = height / weight:

mutate(gsw, ht_wt = height / weight)
## # A tibble: 5 x 4
##           player height weight     ht_wt
##            <chr>  <int>  <int>     <dbl>
## 1 Draymond Green     79    230 0.3434783
## 2   Kevin Durant     81    240 0.3375000
## 3  Klay Thompson     79    215 0.3674419
## 4  Stephen Curry     75    190 0.3947368
## 5  Zaza Pachulia     83    270 0.3074074

In order to permanently change the data, you need to assign the changes to an object:

gsw2 <- mutate(gsw, ht_m = height * 0.0254, wt_kg = weight * 0.4536)
gsw2
## # A tibble: 5 x 5
##           player height weight   ht_m   wt_kg
##            <chr>  <int>  <int>  <dbl>   <dbl>
## 1 Draymond Green     79    230 2.0066 104.328
## 2   Kevin Durant     81    240 2.0574 108.864
## 3  Klay Thompson     79    215 2.0066  97.524
## 4  Stephen Curry     75    190 1.9050  86.184
## 5  Zaza Pachulia     83    270 2.1082 122.472

Reordering rows: arrange()

The next basic verb of "dplyr" is arrange() which allows you to reorder rows. For example, here’s how to arrange the rows of gsw by height

# order rows by height (increasingly)
arrange(gsw, height)
## # A tibble: 5 x 3
##           player height weight
##            <chr>  <int>  <int>
## 1  Stephen Curry     75    190
## 2 Draymond Green     79    230
## 3  Klay Thompson     79    215
## 4   Kevin Durant     81    240
## 5  Zaza Pachulia     83    270

By default arrange() sorts rows in increasing. To arrande rows in descending order you need to use the auxiliary function desc().

# order rows by height (decreasingly)
arrange(gsw, desc(height))
## # A tibble: 5 x 3
##           player height weight
##            <chr>  <int>  <int>
## 1  Zaza Pachulia     83    270
## 2   Kevin Durant     81    240
## 3 Draymond Green     79    230
## 4  Klay Thompson     79    215
## 5  Stephen Curry     75    190
# order rows by height, and then weight
arrange(gsw, height, weight)
## # A tibble: 5 x 3
##           player height weight
##            <chr>  <int>  <int>
## 1  Stephen Curry     75    190
## 2  Klay Thompson     79    215
## 3 Draymond Green     79    230
## 4   Kevin Durant     81    240
## 5  Zaza Pachulia     83    270

Your Turn

  • using the data frame gsw, add a new variable product with the product of height and weight.
gsw <- mutate(gsw, product = height * weight)
gsw
## # A tibble: 5 x 4
##           player height weight product
##            <chr>  <int>  <int>   <int>
## 1 Draymond Green     79    230   18170
## 2   Kevin Durant     81    240   19440
## 3  Klay Thompson     79    215   16985
## 4  Stephen Curry     75    190   14250
## 5  Zaza Pachulia     83    270   22410
  • create a new data frame gsw3, by adding columns log_height and log_weight with the log transformations of height and weight.
gsw3 <- mutate(gsw, log_height = log(height), log_weight = log(weight))
gsw3
## # A tibble: 5 x 6
##           player height weight product log_height log_weight
##            <chr>  <int>  <int>   <int>      <dbl>      <dbl>
## 1 Draymond Green     79    230   18170   4.369448   5.438079
## 2   Kevin Durant     81    240   19440   4.394449   5.480639
## 3  Klay Thompson     79    215   16985   4.369448   5.370638
## 4  Stephen Curry     75    190   14250   4.317488   5.247024
## 5  Zaza Pachulia     83    270   22410   4.418841   5.598422
  • use the original data frame to filter() and arrange() those players with height less than 71 inches tall, in increasing order.
lt71 <- filter(dat, height<71)
arrange(lt71, height)
##          player team position height weight age experience
## 1 Isaiah Thomas  BOS       PG     69    185  27          5
## 2    Kay Felder  CLE       PG     69    176  21          0
## 3    Tyler Ulis  PHO       PG     70    150  21          0
##                    college  salary games minutes points points3 points2
## 1 University of Washington 6587132    76    2569   2199     245     437
## 2       Oakland University  543471    42     386    166       7      55
## 3   University of Kentucky  918369    61    1123    444      21     163
##   points1
## 1     590
## 2      35
## 3      55
  • display the name, team, and salary, of the top-5 highest paid players
salary_sort <- arrange(dat, desc(salary))
top5paid <- slice(salary_sort, 1:5)
top5paid <- select(top5paid, player, team, salary)
top5paid
## # A tibble: 5 x 3
##          player  team   salary
##           <chr> <chr>    <dbl>
## 1  LeBron James   CLE 30963450
## 2    Al Horford   BOS 26540100
## 3 DeMar DeRozan   TOR 26540100
## 4  Kevin Durant   GSW 26540100
## 5  James Harden   HOU 26540100
  • display the name, team, and points3, of the top 10 three-point players
sort3p <- arrange(dat, desc(points3))
top3p <- slice(sort3p, 1:10)
top3p <- select(top3p, player, team, points3)
top3p
## # A tibble: 10 x 3
##            player  team points3
##             <chr> <chr>   <int>
##  1  Stephen Curry   GSW     324
##  2  Klay Thompson   GSW     268
##  3   James Harden   HOU     262
##  4    Eric Gordon   HOU     246
##  5  Isaiah Thomas   BOS     245
##  6   Kemba Walker   CHO     240
##  7   Bradley Beal   WAS     223
##  8 Damian Lillard   POR     214
##  9  Ryan Anderson   HOU     204
## 10    J.J. Redick   LAC     201
  • create a data frame gsw_mpg of GSW players, that contains variables for player name, experience, and min_per_game (minutes per game), sorted by min_per_game (in descending order)
gsw_mpg <- filter(dat, team=="GSW")
gsw_mpg <- mutate(gsw_mpg, min_per_game = minutes / games)
gsw_mpg <- arrange(gsw_mpg, desc(min_per_game))
gsw_mpg <- select(gsw_mpg, player, experience, min_per_game)
gsw_mpg
##                  player experience min_per_game
## 1         Klay Thompson          5    33.961538
## 2         Stephen Curry          7    33.392405
## 3          Kevin Durant          9    33.387097
## 4        Draymond Green          4    32.513158
## 5        Andre Iguodala         12    26.289474
## 6           Matt Barnes         13    20.500000
## 7         Zaza Pachulia         13    18.114286
## 8      Shaun Livingston         11    17.697368
## 9         Patrick McCaw          0    15.126761
## 10            Ian Clark          3    14.766234
## 11           David West         13    12.558824
## 12         JaVale McGee          8     9.597403
## 13 James Michael McAdoo          2     8.788462
## 14         Damian Jones          0     8.500000
## 15         Kevon Looney          1     8.433962

Summarizing values with summarise()

The next verb is summarise(). Conceptually, this involves applying a function on one or more columns, in order to summarize values. This is probably easier to understand with one example.

Say you are interested in calculating the average salary of all NBA players. To do this “a la dplyr” you use summarise(), or its synonym function summarize():

# average salary of NBA players
summarise(dat, avg_salary = mean(salary))
##   avg_salary
## 1    6187014

Calculating an average like this seems a bit verbose, especially when you can directly use mean() like this:

mean(dat$salary)
## [1] 6187014

So let’s make things a bit more interesting. What if you want to calculate some summary statistics for salary: min, median, mean, and max?

# some stats for salary (dplyr)
summarise(
  dat, 
  min = min(salary),
  median = median(salary),
  avg = mean(salary),
  max = max(salary)
)
##    min  median     avg      max
## 1 5145 3500000 6187014 30963450

Well, this may still look like not much. You can do the same in base R (there are actually better ways to do this):

# some stats for salary (base R)
c(min = min(dat$salary), 
  median = median(dat$salary),
  median = mean(dat$salary),
  max = max(dat$salary))
##      min   median   median      max 
##     5145  3500000  6187014 30963450

Grouped operations

To actually appreciate the power of summarise(), we need to introduce the other major basic verb in "dplyr": group_by(). This is the function that allows you to do perform data aggregations, or grouped operations.

Let’s see the combination of summarise() and group_by() to calculate the average salary by team:

# average salary, grouped by team
summarise(
  group_by(dat, team),
  avg_salary = mean(salary)
)
## # A tibble: 30 x 2
##     team avg_salary
##    <chr>      <dbl>
##  1   ATL    6491892
##  2   BOS    6127673
##  3   BRK    4363414
##  4   CHI    6138459
##  5   CHO    6683086
##  6   CLE    8386014
##  7   DAL    6139880
##  8   DEN    5225533
##  9   DET    6871594
## 10   GSW    6579394
## # ... with 20 more rows

Here’s a similar example with the average salary by position:

# average salary, grouped by position
summarise(
  group_by(dat, position),
  avg_salary = mean(salary)
)
## # A tibble: 5 x 2
##   position avg_salary
##      <chr>      <dbl>
## 1        C    6987682
## 2       PF    5890363
## 3       PG    6069029
## 4       SF    6513374
## 5       SG    5535260

Here’s a more fancy example: average weight and height, by position, displayed in desceding order by average height:

arrange(
  summarise(
    group_by(dat, position),
    avg_height = mean(height),
    avg_weight = mean(weight)),
  desc(avg_height)
)
## # A tibble: 5 x 3
##   position avg_height avg_weight
##      <chr>      <dbl>      <dbl>
## 1        C   83.25843   250.7978
## 2       PF   81.50562   235.8539
## 3       SF   79.63855   220.4699
## 4       SG   77.02105   204.7684
## 5       PG   74.30588   188.5765

Your turn:

  • use summarise() to get the largest height value.
summarise(
  dat, 
  max_height = max(height)
)
##   max_height
## 1         87
  • use summarise() to get the standard deviation of points3.
summarise(
  dat, 
  points3_sd = sd(points3)
)
##   points3_sd
## 1    55.9721
  • use summarise() and group_by() to display the median of three-points, by team.
summarise(
  group_by(dat, team),
  mid3p = median(points3)
)
## # A tibble: 30 x 2
##     team mid3p
##    <chr> <dbl>
##  1   ATL  32.5
##  2   BOS  46.0
##  3   BRK  44.0
##  4   CHI  32.0
##  5   CHO  17.0
##  6   CLE  62.0
##  7   DAL  53.0
##  8   DEN  53.0
##  9   DET  28.0
## 10   GSW  18.0
## # ... with 20 more rows
  • display the average triple points by team, in ascending order, of the bottom-5 teams (worst 3pointer teams)
slice(
  arrange(
    summarise(
      group_by(dat, team),
      avg_3p = mean(points3)
    ), 
    avg_3p
  ),
  1:5
)
## # A tibble: 5 x 2
##    team   avg_3p
##   <chr>    <dbl>
## 1   NOP 36.64286
## 2   SAC 37.20000
## 3   PHO 37.60000
## 4   CHI 37.66667
## 5   LAL 39.46667
  • obtain the mean and standard deviation of age, for Power Forwards, with 5 and 10 years (including) years of experience.
pf5and10 = filter(dat, position=="PF" & (experience == 5 | experience==10))
summarise(
  pf5and10,
  avg_age = mean(age),
  sd_age = sd(age)
)
##    avg_age   sd_age
## 1 27.44444 2.351123

First contact with ggplot()

The package "ggplot2" is probably the most popular package in R to create beautiful graphics. In contrast to the functions in the base package "graphcics", the package "ggplot2" follows a somewhat different philosophy, and it tries to be more consistent and modular as possible.

Scatterplots

Let’s start with a scatterplot of salary and points

# scatterplot (option 1)
ggplot(data = dat) +
  geom_point(aes(x = points, y = salary))

  • ggplot() creates an object of class "ggplot"
  • the main input for ggplot() is data which must be a data frame
  • then we use the "+" operator to add a layer
  • the geometric object (geom) are points: geom_points()
  • aes() is used to specify the x and y coordinates, by taking columns points and salary from the data frame

The same scatterplot can also be created with this alternative, and more common use of ggplot()

# scatterplot (option 2)
ggplot(data = dat, aes(x = points, y = salary)) +
  geom_point()

Say you want to color code the points in terms of the position

# colored scatterplot 
ggplot(data = dat, aes(x = points, y = salary)) +
  geom_point(aes(color = position))

Maybe you wan to modify the size of the dots in terms of points3:

# sized and colored scatterplot 
ggplot(data = dat, aes(x = points, y = salary)) +
  geom_point(aes(color = position, size = points3))

To add some transparency effect to the dots, you can use the alpha parameter

# sized and colored scatterplot 
ggplot(data = dat, aes(x = points, y = salary)) +
  geom_point(aes(color = position, size = points3), alpha = 0.7)

Notice that alpha was specified outside aes(). This is because we are not using any column for the alpha transparency values.

Your turn:

  • If you didn’t before, now it’s time to open the ggplot2 cheatsheet
  • Use the data frame gsw to make a scatterplot of height and weight
myplot <- ggplot(data = gsw, aes(x = height, y = weight)) + 
  geom_point()
myplot

- Find out how to make another scatterplot of height and weight, using geom_text() to display the names of the players

?geom_text
myplot + geom_text(aes(label=player))

- Get a scatter plot of height and weight, for ALL the warriors, displaying their names with geom_label()

myplot+geom_label(aes(label=player))

- Get a density plot of salary (for all NBA players)

ggplot(dat, aes(salary)) + geom_density()

- Get a histogram of points2 with binwidth of 50 (for all NBA players)

ggplot(dat, aes(points2)) + geom_histogram()
## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

- Get a barchart of the position frequencies (for all NBA players)

ggplot(dat, aes(position)) + geom_bar()

- Make a scatterplot of experience and salary of all centers, and use geom_smooth() to add a regression line

centers <- filter(dat, position == "C")
cplot <- ggplot(centers, aes(x = experience, y = salary)) + geom_point() 
cplot

- Repeat the same scatterplot of experience and salary of all centers, but now use geom_smooth() to add a loess line

cplot + geom_smooth()
## `geom_smooth()` using method = 'loess'

Faceting

One of the most attractive features of "ggplot2" is the ability to display multiple facets. The idea of facets is to divide a plot into subplots based on the values of one or more categorical (or discrete) variables.

Here’s an example. What if you want to get scatterplots of points and salary separated (or grouped) by position? This is where faceting comes handy, and you can use facet_warp() for this purpose:

# scatterplot by position
ggplot(data = dat, aes(x = points, y = salary)) +
  geom_point() +
  facet_wrap(~ position)

The other faceting function is facet_grid(), which allows you to control the layout of the facets (by rows, by columns, etc)

# scatterplot by position
ggplot(data = dat, aes(x = points, y = salary)) +
  geom_point(aes(color = position), alpha = 0.7) +
  facet_grid(~ position) +
  geom_smooth(method = loess)

# scatterplot by position
ggplot(data = dat, aes(x = points, y = salary)) +
  geom_point(aes(color = position), alpha = 0.7) +
  facet_grid(position ~ .) +
  geom_smooth(method = loess)

Your turn:

  • Make scatterplots of experience and salary faceting by position
ggplot(dat, aes(x = experience, y = salary)) + geom_point() + facet_wrap(~position)

- Make scatterplots of experience and salary faceting by team

ggplot(dat, aes(experience, salary)) + geom_point() + facet_wrap(~team)

- Make density plots of age faceting by team

ggplot(dat, aes(age)) + geom_density() + facet_wrap(~team)

- Make scatterplots of height and weight faceting by position

ggplot(dat, aes(height, weight)) + geom_point() + facet_wrap(~position)

- Make scatterplots of height and weight, with a 2-dimensional density, geom_density2d(), faceting by position

ggplot(dat, aes(height, weight)) + geom_density2d() + facet_wrap(~position)

- Make a scatterplot of experience and salary for the Warriors, but this time add a layer with theme_bw() to get a simpler background

ggplot(dat, aes(experience, salary)) + geom_point() + theme_bw()

- Repeat any of the previous plots but now adding a leyer with another theme e.g. theme_minimal(), theme_dark(), theme_classic()

ggplot(dat, aes(experience, salary)) + geom_point() + theme_minimal()

ggplot(dat, aes(experience, salary)) + geom_point() + theme_dark()

ggplot(dat, aes(experience, salary)) + geom_point() + theme_classic()